The electrification of high-performance motorsport demands power conversion systems that simultaneously achieve exceptional power density, robust electromagnetic behavior, and rigorous functional safety. This thesis describes the evolution and refinement of a Silicon Carbide (SiC) MOSFET-based three-phase inverter conceived for the four-wheel-drive electric powertrain of the Dynamis PRC Formula Student vehicle at Politecnico di Milano, and advances the platform along three distinct engineering axes. The first contribution addresses electromagnetic compatibility. The previous prototype was constrained to 9000 rpm by noise coupling into the rotor position sensor readout, which corrupted the sin/cos differential signals and destabilized the field-oriented control loop. A systematic PCB redesign was carried out: ground-plane domains were placed by signal type, split-termination networks were applied to the differential signal lines, and routing was revised to enforce geometric balance throughout the sensor signal path. The interventions reduced RMS noise on the sin channel by 14%, extending reliable motor operation to 18000 rpm, with the residual gap to the 20000 rpm motor ceiling attributed to the field-weakening algorithm rather than signal integrity. The second contribution is a functional safety software framework built around a deterministic Finite State Machine (FSM) derived from the CiA 402 drive profile and tailored to Formula Student requirements. The FSM introduces two fault severity levels, Hard Fault and Recoverable Fault, to prevent unnecessary tractive system shutdowns in response to transient anomalies. A precharge Built-In Self-Test (BIST) executes before the main positive contactor closes: individual legs are driven sequentially to verify gate driver and power modules integrity, followed by symmetric H-bridge excitation at 50% duty cycle to detect phase disconnections and validate current sensor readings under zero-net-torque conditions. At runtime, a two-level sin/cos consistency monitor computes the Pythagorean norm of the normalized encoder signals at 16 kHz for hard fault detection, while a 100 Hz background monitor tracks running extrema to flag gain mismatch and DC offset. The fast check requires at most 1.33 us per control period, representing a negligible computational overhead on the STM32G4 microcontroller. The third contribution is an averaged power losses model for the SiC inverter. An analytical IGBT-based averaging framework, which operates at the timescale of the motor control loop rather than at the switching frequency, was adapted to SiC MOSFET technology by removing the bipolar threshold voltage from conduction losses expressions, restricting body diode conduction to dead-time intervals, and applying a scalar scaling for the negligible reverse recovery energy. Validation against measurements from an HBK power analyzer in a back-to-back motor test revealed a systematic underprediction, which was traced to resistive contributions external to the MOSFET channel — PCB copper paths, press-fit contacts, and phase cable segments — totaling approximately 12.6 mOhm per phase. Including this term brought model predictions into agreement with experimental data, confirming the validity of the modeling approach and its suitability for integration into a vehicle complex dynamic model, to enhance powertrain efficiency and thermal management analysis.
La transizione verso propulsori elettrici ad alte prestazioni nel motorsport richiede sistemi di conversione di potenza capaci di coniugare elevata densità di potenza, robustezza elettromagnetica e sicurezza funzionale rigorosa. Questa tesi presenta l'evoluzione e il miglioramento di un inverter trifase basato su MOSFET al Carburo di Silicio (SiC), progettato per il powertrain elettrico del veicolo del Dynamis PRC, la squadra di Formula Student del Politecnico di Milano. Il lavoro si articola su tre contributi principali. Il primo riguarda la compatibilità elettromagnetica. Il prototipo precedente era limitato a 9000 giri/min a causa dell'accoppiamento del rumore nella catena di lettura del sensore di posizione del rotore, che corrompeva i segnali differenziali sin/cos e destabilizzava l'algoritmo di Field-Oriented Control. È stata condotta una riprogettazione sistematica del PCB: i piani di massa sono stati riposizionati in base al dominio dei segnali, sono state applicate reti di terminazione split alle linee differenziali e il routing è stato rivisto per garantire la simmetria geometrica dell'intero percorso del segnale. Gli interventi hanno ridotto il rumore RMS sul canale sinusoidale del 14%, estendendo il funzionamento stabile del motore fino a 18000 giri/min. Il divario residuo rispetto al limite fisico di 20000 giri/min è attribuibile all'algoritmo di field-weakening e non all'integrità del segnale. Il secondo contributo riguarda la sicurezza funzionale del software. È stato sviluppato un framework strutturato attorno a una Macchina a Stati Finiti (FSM) derivata dal profilo CiA 402 e adattata al contesto Formula Student. La FSM introduce due livelli di severità del guasto, Hard Fault e Recoverable Fault, per evitare spegnimenti non necessari del sistema di trazione in risposta ad anomalie transitorie. Un Built-In Self-Test (BIST) di precarica viene eseguito prima della chiusura del contattore positivo principale: ogni gamba dell'inverter viene eccitata individualmente per verificare l'integrità del gate driver e dei moduli di potenza, seguita da un'eccitazione simmetrica in configurazione a ponte H al 50% di duty cycle, per individuare disconnessioni di fase e validare le letture dei sensori di corrente in condizioni di coppia netta nulla. A regime, un controllo a due livelli della consistenza sin/cos calcola la norma pitagorica dei segnali dell'encoder normalizzati a 16 kHz per il rilevamento di guasti gravi, mentre un monitor di fondo a 100 Hz traccia i valori estremi per segnalare diffenze di guadagno e offset. Il controllo rapido richiede al massimo 1.33 us per periodo di controllo, con un overhead computazionale trascurabile sul microcontrollore STM32G4. Il terzo contributo è un modello analitico per la stima mediata delle perdite di potenza dell'inverter. Si tratta di un modello originariamente sviluppato per inverter a tecnologia IGBT, che opera alla scala temporale del ciclo di controllo del motore anziché alla frequenza di commutazione, che è stato adattato ai MOSFET SiC eliminando la tensione di soglia bipolare dalle espressioni di perdita in conduzione, limitando la conduzione del diodo di body agli intervalli di dead-time e introducendo uno scaling per l'energia di recupero inverso, trascurabile nei dispositivi SiC. La validazione tramite analizzatore di potenza HBK in una configurazione back-to-back ha evidenziato una sottostima sistematica, ricondotta a resistenze parassite esterne al canale MOSFET, ovvero a percorsi in rame sul PCB, contatti press-fit e tratti di cavo di fase, pari complessivamente a circa 12.6 mOhm per fase. Includendo questo contributo, le previsioni del modello sono risultate in accordo con i dati sperimentali, confermandone la validità e l'idoneità all'integrazione in un modello dinamico completo del veicolo per l'analisi dell'efficienza del powertrain e della gestione termica.
Motorsport SiC-based inverter: EMC-oriented hardware design, functional safety software implementation and averaged losses validation
FERRARI, LAURA
2025/2026
Abstract
The electrification of high-performance motorsport demands power conversion systems that simultaneously achieve exceptional power density, robust electromagnetic behavior, and rigorous functional safety. This thesis describes the evolution and refinement of a Silicon Carbide (SiC) MOSFET-based three-phase inverter conceived for the four-wheel-drive electric powertrain of the Dynamis PRC Formula Student vehicle at Politecnico di Milano, and advances the platform along three distinct engineering axes. The first contribution addresses electromagnetic compatibility. The previous prototype was constrained to 9000 rpm by noise coupling into the rotor position sensor readout, which corrupted the sin/cos differential signals and destabilized the field-oriented control loop. A systematic PCB redesign was carried out: ground-plane domains were placed by signal type, split-termination networks were applied to the differential signal lines, and routing was revised to enforce geometric balance throughout the sensor signal path. The interventions reduced RMS noise on the sin channel by 14%, extending reliable motor operation to 18000 rpm, with the residual gap to the 20000 rpm motor ceiling attributed to the field-weakening algorithm rather than signal integrity. The second contribution is a functional safety software framework built around a deterministic Finite State Machine (FSM) derived from the CiA 402 drive profile and tailored to Formula Student requirements. The FSM introduces two fault severity levels, Hard Fault and Recoverable Fault, to prevent unnecessary tractive system shutdowns in response to transient anomalies. A precharge Built-In Self-Test (BIST) executes before the main positive contactor closes: individual legs are driven sequentially to verify gate driver and power modules integrity, followed by symmetric H-bridge excitation at 50% duty cycle to detect phase disconnections and validate current sensor readings under zero-net-torque conditions. At runtime, a two-level sin/cos consistency monitor computes the Pythagorean norm of the normalized encoder signals at 16 kHz for hard fault detection, while a 100 Hz background monitor tracks running extrema to flag gain mismatch and DC offset. The fast check requires at most 1.33 us per control period, representing a negligible computational overhead on the STM32G4 microcontroller. The third contribution is an averaged power losses model for the SiC inverter. An analytical IGBT-based averaging framework, which operates at the timescale of the motor control loop rather than at the switching frequency, was adapted to SiC MOSFET technology by removing the bipolar threshold voltage from conduction losses expressions, restricting body diode conduction to dead-time intervals, and applying a scalar scaling for the negligible reverse recovery energy. Validation against measurements from an HBK power analyzer in a back-to-back motor test revealed a systematic underprediction, which was traced to resistive contributions external to the MOSFET channel — PCB copper paths, press-fit contacts, and phase cable segments — totaling approximately 12.6 mOhm per phase. Including this term brought model predictions into agreement with experimental data, confirming the validity of the modeling approach and its suitability for integration into a vehicle complex dynamic model, to enhance powertrain efficiency and thermal management analysis.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_7_Ferrari_Executive Summary.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Executive summary
Dimensione
2.35 MB
Formato
Adobe PDF
|
2.35 MB | Adobe PDF | Visualizza/Apri |
|
2026_7_Ferrari_Tesi.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Tesi
Dimensione
14.73 MB
Formato
Adobe PDF
|
14.73 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/261464